Spend two days in these code,very stupid.
Note the important part ‘%must set rows 1 = 0,important!!! ‘.
The code about gradient descent should be computed by matrix.
It will be very quickly in running.
function [J grad] = nnCostFunction(nn_params, ...
input_layer_size, ...
hidden_layer_size, ...
num_labels, ...
X, y, lambda)
%NNCOSTFUNCTION Implements the neural network cost function for a two layer
%neural network which performs classification
% [J grad] = NNCOSTFUNCTON(nn_params, hidden_layer_size, num_labels, ...
% X, y, lambda) computes the cost and gradient of the neural network. The
% parameters for the neural network are "unrolled" into the vector
% nn_params and need to be converted back into the weight matrices.
%
% The returned parameter grad should be a "unrolled" vector of the
% partial derivatives of the neural network.
%
% Reshape nn_params back into the parameters Theta1 and Theta2, the weight matrices
% for our 2 layer neural network
%Theta1 = (25*401)
Theta1 = reshape(nn_params(1:hidden_layer_size * (input_layer_size + 1)), ...
hidden_layer_size, (input_layer_size + 1));
%Theta2 = (10*26)
Theta2 = reshape(nn_params((1 + (hidden_layer_size * (input_layer_size + 1))):end), ...
num_labels, (hidden_layer_size + 1));
% Setup some useful variables
m = size(X, 1);
% You need to return the following variables correctly
J = 0;
Theta1_grad = zeros(size(Theta1));
Theta2_grad = zeros(size(Theta2));
% ====================== YOUR CODE HERE ======================
% Instructions: You should complete the code by working through the
% following parts.
%
% Part 1: Feedforward the neural network and return the cost in the
% variable J. After implementing Part 1, you can verify that your
% cost function computation is correct by verifying the cost
% computed in ex4.m
%
% 1.3 feedforeward and cost function
%y_ = zeros(m,num_labels); %onehot y_ = (5000*10)
%for i=1:m
% y_(i,y(i))=1;
%endfor
y_ = eye(num_labels)(y,:);
X = [ones(m,1) X]; %X = (5000*401)
z2 = X * Theta1'; %z2 = (5000*25)
a2 = sigmoid(z2); %a2 = (5000*25)
a2_addone = [ones(m,1) a2]; %a2_addone =(5000*26)
z3 = a2_addone * Theta2'; %z3= (5000*10)
a3 = sigmoid(z3); %a3= (5000*10)
%temp=(5000*10)
temp = (-1*y_).*log(a3) - (1-y_).*log(1-a3); %����ʹ�õ��ˣ�
J = sum(sum(temp))/m;
% 1.4 Regularized Cost Function
n1 = size(Theta1,2);
t1 = Theta1(:,2:n1); %t1 = (5000*400)
n2 = size(Theta2,2);
t2 = Theta2(:,2:n2);
J = J + (lambda/(2*m))*(sum(sum(t1.^2))+sum(sum(t2.^2)));
% Part 2: Implement the backpropagation algorithm to compute the gradients
% Theta1_grad and Theta2_grad. You should return the partial derivatives of
% the cost function with respect to Theta1 and Theta2 in Theta1_grad and
% Theta2_grad, respectively. After implementing Part 2, you can check
% that your implementation is correct by running checkNNGradients
%
% Note: The vector y passed into the function is a vector of labels
% containing values from 1..K. You need to map this vector into a
% binary vector of 1's and 0's to be used with the neural network
% cost function.
%
% Hint: We recommend implementing backpropagation using a for-loop
% over the training examples if you are implementing it for the
% first time.
%Theta1 = (25*401)
%Theta2 = (10*26)
DELTA_1 = zeros(hidden_layer_size,(input_layer_size+1)); %(25*401)
DELTA_2 = zeros(num_labels,(hidden_layer_size+1)); %(10*26)
Theta1_regulization = Theta1*lambda/m; %(25*401)
Theta1_regulization(:,1) = 0; % columns 1 need to set zeros.
Theta2_regulization = Theta2*lambda/m; %(10*26)
Theta2_regulization(:,1) = 0; % columns 1 need to set zeros.
for t = 1:m
a_1 = X(t,:); %a_1 = (1*401)
z_2 = Theta1*a_1'; %z_2 = (25*1)
a_2 = sigmoid(z_2);
a_2 = [1;a_2]; %a_2 = (26*1)
z_3 = Theta2 * a_2; %z_3 = (10*1)
a_3 = sigmoid(z_3); %a_3 = (10*1)
delta_3 = a_3 - y_(t,:)'; %delta_3 = (10*1)
delta_2 = (Theta2'*delta_3).* sigmoidGradient([1;z_2]); %delta_2 = (26*1)
delta_2(1,:)=1; %must set rows 1 = 0,important!!!
%delta_2(:,2:end) = (25*1)
DELTA_1 = DELTA_1 + delta_2(2:end) * a_1 ; %(25*401)
%a_2(2:end) = (25*1)
DELTA_2 = DELTA_2 + delta_3 * a_2'; %(10*26)
endfor
Theta1_grad = DELTA_1/m + Theta1_regulization;
Theta2_grad = DELTA_2/m + Theta2_regulization;
grad = [Theta1_grad(:);Theta2_grad(:)];
%
% Part 3: Implement regularization with the cost function and gradients.
%
% Hint: You can implement this around the code for
% backpropagation. That is, you can compute the gradients for
% the regularization separately and then add them to Theta1_grad
% and Theta2_grad from Part 2.
%
% -------------------------------------------------------------
% =========================================================================
% Unroll gradients
grad = [Theta1_grad(:) ; Theta2_grad(:)];
end